A fat

FRUSTRATION AND AGGRESSION (F-A) THEORY Johan M.G. van der Dennen In 1939, Dollard, Doob, Miller, Mowrer, and Sears published a monograph on aggression in which they presented what has come to be known as the frustration-aggression hypothesis (F-A). This hypothesis proved to have an immense impact. It appears to have influenced current Western thinking on aggression more profoundly than any other single publication.

For more than three decades, the F-A hypothesis has guided, in one way or another, the better part of the experimental research on human aggression (Cf. Geen, 1972; Zillmann, 1979).

Perhaps more importantly, however, the views of aggression that it involves seem to have become widely adopted and accepted; they have become commonplaces. This popular success may have various sources. First the principal hypothesis is uncomplicated and easy to grasp.

The theory is generally well structured and clearly articulated, a fact that again facilitates comprehension. Second, the theory does not involve overly abstract concepts or elaborate procedures. It is very close to common sense – seeming to be built on it. Finally, as Selg (1971) observed, the theory tends to provide a justification for behaving aggressively: ‘Being frustrated made me do it!’ Like the aggression amnesty provided by instinct notions (‘It can’t be helped because we’re built that way’), although not as strong, this kind of justification can be drawn upon as a ready-made excuse for uncontrolled (or premeditated) hostile or aggressive actions (Zillmann, 1979).

As to the principal hypothesis, Dollard et al. (1939) posited “that the occurrence of aggressive behavior always presupposes the existence of frustration and, contrariwise, that the existence of frustration always leads to some form of aggression”. Frustration, in this context, was specified as the thwarting of a goal response, and a goal response, in turn, was taken to mean the reinforcing final operation in an ongoing behavior sequence. At times, however, the term ‘frustration’ is used to refer not only to the process of blocking a person’s attainment of a reinforcer but also to the reaction to such blocking. Consequently, ‘being frustrated’ means both that one’s access to reinforcers is being thwarted by another party (or possibly by particular circumstances) and that one’s reaction to this thwarting is one of annoyance.

It was soon recognized that the initial claims – (a) that aggression is always based on frustration and (b) that frustration always leads to aggression – were far too general. These blaims made frustration both a necessary and sufficient condition for aggression. Miller (1941) was quick to retract the latter part of the proposal. Quite obviously, frustrations do not cause hostile or aggressive outbursts by necessity. Potential outbursts may be effectively inhibited or may result in alternative actions, such as the pursuit of other, more readily available reinforcers. Miller therefore rephrased the second part of the hypothesis to read: “Frustration produces instigations to a number of different types of response, one of which is an instigation to some form of aggression”.

According to this reformulation, frustration actuates motivational forces that are diffuse rather than specific to aggression. It is assigned the properties of a general drive. Such apparent moderation has not been applied to the first part of the original F-A hypothesis, however.

Miller (1941) found the generality of this claim both defensible and useful.

The revised F-A hypothesis thus maintains the following: (a) Frustration instigates behavior that may or may not be hostile or aggressive. (b) Any hostile or aggressive behavior that occurs is caused by frustration. In other words, frustration is not a sufficient, but a necessary, condition for hostility and aggression (Zillmann, 1979).

It should be noted that the revised hypothesis retains a good deal of the original, sweeping claim. Because of its sweeping nature, the hypothesis proved most controversial (cf. Bandura

& Walters, 1963a,b; Buss, 1961). After considering the more specific elements of frustration-aggression theory, we briefly review the main arguments in this controversy.

In developing a comprehensive theory of aggression, Dollard et al. (1939) specified that the motivational strength toward aggression is a function of: (a) the reinfocement value of the frustrated goal response, (b) the degree of frustration of this goal response, and (c) the number of frustrated response sequences. The first two of these propositions are straightforward.

Aggression-potentiating annoyance is seen to increase with the incentive that could be obtained or the aversion that could be terminated by the blocked goal reaction. Furthermore, frustration can be incomplete, and thus a goal reaction can be partially completed. The third proposition is less direct, however. It is meaningful only if it is assumed that frustration-induced annoyance is cumulative. It is apparently held that ‘aggressive drive’ resulting from frustrations is somehow maintained within the organism and adds up to a level at which an otherwise tolerable frustration evokes aggression. Dollard et al. were, in fact, very explicit about the assumed additivity of aggressive forces. They posited that the strength of a hostile or aggressive reaction depends in part on the “amount of residual instigation from previous or simultaneous frustrations”. “Minor frustrations” they suggested, “add together to produce an aggressive response of greater strength than would normally be expected from the frustrating situation that appears to be the immediate antecedent of the aggression”. Dollard et al. acknowledged the significance of the temporal aspect of this summation of ‘aggressive drive’ but quickly dismissed the issue by pointing out the lack of relevant data (Zillmann, 1979).

The theoretical treatment of the inhibition of aggression is related to the time issue, in that the lack of immediate, overt manifestations of aggression is assumed to lead to prolonged covert consequences that eventually ‘break out’ in different form. Dollard et al. recognized that not all frustrations produce overt aggression, and to account for this fact, they posited inhibitory forces whose strength was said to vary positively with the severity of the punishment anticipated to result from the particular contemplated goal reaction. It was proposed that if punishment (a notion that was broadened to include such things as injury to a loved object and failure to achieve desired objectives) was anticipated to outweigh any incentives that could be gained, overt aggression would be inhibited. However, consistent with the original conviction that all frustrations produce some form of aggression, Dollard et a1. insisted that it would be ‘clearly false’ to view inhibited overt aggression as nonaggression. Being ‘furious inside’, for example, is interpreted as nonovert aggression, which apparently can linger on and erupt in overt manifestations at a later time.

Put precisely, then, anticipated punishment, which is a primary source of frustration, effects the inhibition of overt aggression when it exceeds anticipated gratifications. The inhibition is incomplete, however, in the sense that: (a) it fails to control covert elements of aggression, and (b) it fails to terminate the instigation to aggression.

A significant element of frustration-aggression theory concerns the redirection or displacement of aggression. Dollard et al. were very explicit in their treatment of this phenomenon. They proposed that a particular frustration instigates aggression primarily against the source of the frustration but also instigates aggression against targets that are to some degree related to that source. The strength of the instigation was seen to vary as a function of associative ties between the actual source of frustration and the a1ternative target.

The Freudian displacement mechanism was thus interpreted in terms of stimulus affinities.

With the source of frustration constituting the primary target for aggression, closely associated targets evoke similar aggressive reactions; generally, the strength of the instigation to aggression diminishes as the similarity between the original and alternative target decreases. Dollard et al. further proposed that the more punishment is anticipated to follow contemplated acts of aggression against a particular target – that is, the more severe the

inhibition placed upon such behavior – the more likely it becomes that the ‘inhibited’ aggressive actions will be: (a} replaced by alternative, less punishment-burdened acts, and/or (b) displaced upon other targets. These propositions make it clear once again that in the framework of frustration-aggression theory, the inhibition of aggression is always incomplete.

The frustrated individual who is forced to inhibit contemplated acts against particular targets (this inhibition in itself being considered frustrating and thus increasing the frustration suffered) is viewed as motivated to find other outlets for his or her aggressive inclinations.

Only hostile or aggressive activities, transformed or displaced as they may be, are capable of reducing this instigation to aggression. Since the theory fails to stipulate other mechanisms for the reduction of such instigation, frustration must be viewed as a force that ‘drives’ the organism for an indefinite period of time, ultimately until hostile and aggressive acts are performed (Zillmann, 1979).

This state of affairs is not entirely consistent with Miller’s (1941) revision of the F-A hypothesis. The suggestion that frustration instigates nonaggressive as well as aggressive reactions seems to imply the possibility of instigation reduction by nonaggressive goal reactions.

Concerning the reduction of the instigation to aggression as such, Dollard et al. (1939) categorically declared that it is achieved, at least in part, with any and every act of aggression.

These investigators further proposed an ‘equivalence of forms’ of instigation-reducing means, positing an “inverse relationship between the occurrence of different forms of aggression”.

They qualified this postulated reciprocal relationship by suggesting that it applies especially to the dichotomies of overt vs. covert and self-directed vs. outward-directed aggression. In this context, the notion of catharsis is equated with the reduction of the instigation to aggression in general, irrespective of specific targets.

A violent assault upon a frustrator is thus seen as cathartic. More significantly, however, the expression of minor, less direct, and possibly covert acts of ‘aggression’ are viewed as alternative, powerful means to bring about catharsis. The mere expression of annoyance, which does not harm anybody, is also considered an aggressive act capable of producing catharsis. Furthermore (and very significantly), since according to the theory, aggression can be displaced, attacks upon alternative targets are seen to reduce the instigation to aggression against the actual frustrator. Finally, aggression against the self, in overt or covert form, may prove cathartic as well. (In contrast to Freudian thought regarding the death instinct, Dollard et al. consider self-aggression a last resort, however. “Other conditions being constant, self-aggression should be a relatively nonpreferred type of expression which will not occur unless other forms of expression are even more strongly inhibited”. This proposition leaves no doubt that aggression as a reaction to frustration is primarily self-assertive).

In frustration-aggression theory, the treatment of catharsis principally parallels that of the instigation to aggression: Whereas instigation is conceived of as cumulative, with various independent frustrations heightening its level successively, catharsis is cumulative in the sense that independent cathartic incidents successively lower the level of instigation. ‘Drive’ is built up by frustrations and ‘worn down’ by catharsis. According to Dollard et al., this frustration-induced drive is aggression specific without being specific to particular aggressive forms, and it is not specific to particular targets. Miller’s revision of F-A theory affects only the first part of this assessment: Frustration-induced drive is conceived of as a nonspecific motivating force. The two remaining, outstanding features of the theory have been retained.

Zillmann (1979) refers to these two features as: (a) the interchangeability of aggressive forms and (b) the interchangeability of targets.

As noted by various investigators (e.g. Buss, 1961; Selg, 1971), the treatment of the temporal component of both instigation and catharsis in F-A theory is vague and ambiguous. How long

can the instigational effect of frustrations ensue? How long does catharsis exert its instigation-reducing impact?

Consistent with Dollard et al., the many denials of gratifications suffered during childhood can accumulate to intense juvenile frustration, and the struggle of the working class can build up to levels at which it incites revolution. In disadvantaged minorities, the instigation to aggression grows with every frustration until it ultimately leads to violence. These are typically cited long-term effects. The proposed additivity of instigational forces is meaningful only if it is assumed that individual frustrations have extremely long-lasting consequences. In terms of aggressive drive, this means that the drive state, once activated, remains active within the organism for an indefinite length of time. Under these circumstances, drive is a purely hypothetical construct without physical embodiment in the organism. The construct of instigation to aggression, which Dollard et al. preferred, is similarly without embodiment but can, as a motivational force, be approached empirically through self-reports. However, whether the concept of aggressive drive or aggressive instigation is employed, the long-term effects of frustration are without identifiable organismic changes; they are – if at all measurable – difficult to trace, and they have opened the door for considerable speculation.

For example, Plack (1969) derives virtually all the evils of society, especially violence, from the abundance of denials in childhood and early adolescence. Since a good portion of these denials are likely to be employed even in societies far more permissive than ours (the point being that some will have to be employed), such interpretations, which are principally founded on the presumed accumulation of frustrations with long-term instigational effects, come very close to the projection of instinctive aggression. If aggressive drive is assumed to accumulate steadily with the unavoidable frustrations of everyday life, its consequences become indistinguishable from those of a spontaneous instinctive type of aggression. In this connection, it is of little moment that aggression is reactive in F-A theory (Zillmann, 1979).

All this is not to say that extended periods of deprivation from valued objects and experiences cannot promote hostility and aggression. Bandura & Walters (1959), for example, presented evidence showing that aggressive juveniles tended to have suffered a greater incidence of frustration during childhood than nonaggressive ones. At a more extreme level of aggression, Palmer (1980) has suggested that murder can result from frustrations inflicted by illness and accidents. Generally speaking, the belief that nagging frustrations, resulting mainly from the ever-present societally imposed pressures to inhibit relatively minor activities, can eventuate in violent crime in ‘overcontrolled’ persons has grown popular among psychiatrists and lawyers (Cf. Lunde, 1975; Megargee, 1966, 1971). The issue at hand, however, is whether or not observed violent transgressions occur because of the accumulation of residual instigatory effects of frustrations. The aggressive behavior of frustrated people is not by necessity the result of their frustrations, and the account of their actions on the basis of frustrations is usually open to alternative explanations. Violent youths, for example, may behave aggressively, not because of remote childhood frustrations, but because they have developed aggressive dispositions through the modeling of their typically more aggressive parents and because they have become to experience less intense guilt feelings regarding aggression (Bandura & Walters, 1959). Similarly, crippling illnesses or accidents may cause violent eruptions not so much because instigational forces are steadily built up but because the person so affected, as a result of the handicap, is denied gratifications he or she once obtained readily. The person may be largely deprived of his or her earlier means of nonaggressive appeal and resort to violence as the only effective means left to change the behavior of others as desired. Even in ‘overcontrollers’ aggressive outbursts may come after the realization that nonaggressive ways have failed to achieve vital objectives. The roundabout hint at frustrations as the ultimate source of all violence, epitomized in Berkowitz’s (1962) contention that “the person who kills generally does so because he has been frustrated”, is far

too simplistic to be acceptable as an explanatory account. Such a formula merely rephrases the necessity postulate of frustration-aggression theory (i.e., frustration is a necessary – though not a sufficient – condition for aggression), which not only fails to be enlightening but is also erroneous (cf. Buss, 1961, 1971).

The ambiguities concerning the time course of the instigation to aggression following a frustration have plagued F-A theory in that they have fostered far-reaching and farfetched interpretations and applications. Although the theory is equally vague about the time course of cathartic effects, this has not resulted in similarly broad or farfetched interpretations.

Dollard et al. (1939) suggested that catharsis is generally of short duration, with likely frustrations soon again building up the instigation to aggression. This suggestion, it seems, has been universally accepted. According to common interpretations, catharsis at best brings aggressive drive to a zero level. It cannot accumulate a negative ‘counter-force’ that could hold instigational forces below zero. However, consistent with the theory, repeated instances of catharsis should lead to successive decrements in the prevailing instigation to aggression.

Catharsis should thus be as additive and cumulative in lowering instigational forces as frustration is in elevating them. Also, the mechanism assumed to govern the posited sustenance of instigational forces following frustration should equally affect their sustenance after catharsis. That is to say that if frustration effects are long-lasting, cathartic effects should be long-lasting as well. Interestingly, these implications of the theory have been largely neglected. In contrast to the specific recommendations that have been derived from Lorenz’s instinct theory of aggression, in F-A theory proper, it has not been suggested that violence in society can be curbed effectively by a battery of minor catharses; and it has not been claimed that extreme acts of violence – for example, the murder of an annoyer – could induce supercatharses that would cause extended periods of nonhostile, nonaggressive behavior in individuals. Only the somewhat careless interpretation of F-A theory, it seems, has led to the occasional projection of such and similar cathartic effects (Zillmann, 1979).

Restrictions of the Frustration-Aggression Theory Although the F-A hypothesis initially was widely accepted without much modification (e.g., Berkowitz, 1958; McNeil, 1959), there were early critics of the universal claims expressed in it. Both Maslow (1941) and Rosenzweig (1944), for example, suggested that frustrations instigate aggression only when they are associated with threat. More recently, Buss (1961) similarly insisted that the thwarting of a goal reaction in and of itself does not instigate aggression and that in order to evoke aggressive behavior, frustration must involve the element of attack. Buss also stressed the instrumental value of aggression in the overcoming of frustrations. The initial use of the concept of frustration was thus considered to confound thwarting and attack. Berkowitz (1958), for example, initially subsumed insult and attack under the general heading of frustration, arguing that the various interlocking components could not be separated operationally. Subsequent research showed, however, that the factors in question could be isolated both conceptually and operationally. This research has been competently reviewed by several investigators (e.g., Bandura, 1973a; Buss, 1961). On the basis of his own experiments, Buss (1966b) concluded that “frustration is at best a weak determiner of aggression”.

The point that frustration does not necessarily lead to aggression has been made in numerous investigations. In an early experiment, Davitz (1952) observed that the response to frustration varies greatly with the prior training and personality characteristics of the respondent. A recent, more complex study by Christy, Gelfand & Hartmann (1971) makes essentially the

same point. These and many other experiments (Cf. Bandura, 1969, 1973a,b) suggest that aggression is just one of potentially many coping reactions to frustration. After an extensive review of the pertinent research literature, Bandura (1969) concluded that “when thwarted, some people become dependent and seek help and support, some display withdrawal and resignation, some experience psychosomatic dysfunctions, some seek refuge in drug-induced experiences and anaesthetic doses of alcohol, some respond aggressively and most simply intensify constructive efforts to overcome the obstacles they face”. The fact that some studies reported findings supportive of the frustration-aggression connection – that is, increased aggressiveness after frustration (e.g., Buss, 1963; Geen, 1968; Rule & Percival, 1971) – is probably due to the involvement of supplementary factors such as personal attack or the instrumental value of aggressive reactions. Perhaps equally important, the procedural restriction to aggressive reactions in these studies – that is, the elimination of potentially preferred alternative responses – may have led to the observed increments in aggressive reactivity (Zillmann, 1979).

There is considerable evidence supporting the view that frustration becomes a potent inducer of aggression when it is associated with personal attack. In a number of investigations it has been shown that frustration produces hostility or aggression more readily when it is perceived as arbitrary, unjustified, intentional, or unmitigated rather than accidental or warranted (e.g., Burnstein & Worchel, 1962; Cohen, 1955; Kregarman & Worchel, 1961; Mallick & McCandless, 1966; Pastore, 1952; Rothaus & Worchel, 1960; Zillmann et al., 1975; Zillmann & Cantor, 1976b). These investigations generally corroborate Buss’ (1961, 1971) contention that personal attack rather than the blocking of a goal response per se is the principal aggression-potentiating element in frustration – when the popular, broad meaning of this concept is used. The investigations also point up the importance of cognitive considerations in the control of hostility and aggression. For example, Mallick & McCandless (1966) showed that retaliatory behavior is reduced when a reasonable explanation for the provoker’s misbehavior is supplied subsequent to provocation, and Zillmann & Cantor (1976b) demonstrated that prior knowledge of mitigating circumstances associated with a provocation can prevent the development of anger, thereby holding down the level of retaliatory hostilities. In summary, research on the frustration-aggression hypothesis leaves no doubt about the indefensibility of the original proposal that presented frustration as both a necessary and a sufficient condition for aggression (Zillmann, 1979). In light of the relevant research findings, the following restrictive propositions concerning the relationship between frustration and aggression seem to be in order: (1) The blockage of a goal reaction, in and of itself, may prompt the expression of an affective reaction such as disappointment, annoyance, or anger, but it generally will not induce interpersonal hostility or aggression.

(2) The blockage of a goal reaction is likely to evoke behaviors through which the blockage can be terminated and the initial goal reaction executed. If aggressive responses hold promise for such an overcoming of the frustration, their execution becomes likely, particularly in the absence of undesirable repercussions. If aggressive responses have proved in the past to have instrumental value in the overcoming of frustrations, aggressive acts (like nonaggressive, instrumental actions) will acquire habit strength, and the probability of their execution in the face of frustration will increase. Thus, hostile and aggressive reactions in response to the blockage of goal response are likely only if the specified instrumental value of such reactions exceeds that of nonhostile and nonaggressive alternatives or if this condition has prevailed in the past with some consistency.

(3) When coupled with personal attack, goal thwarting is likely to evoke feelings of annoyance and anger, which in turn increase the likelihood of hostile and aggressive reactions. In this context, goal thwarting that the frustrated person perseives as intentional, arbitrary, and unjustified is conceived of as personal attack (Zillmann, 1979).

Extensions of Frustration-Aggression Theory The application of Miller’s (1944) conflict model to aggressive behavior, specifically to the displacement of aggression onto substitute targets, proved to be a highly influential extension of frustration-aggression theory. Miller (1948, 1959) proposed that both the substitute target and the intensity of the attack directed against it could be predicted on the basis of three antecedent conditions: (a) the strength of the drive that motivates aggression against the original target, (b) the strength of inhibitory response tendencies, and (c) the degree of stimulus similarity between the original and substitute targets. More specifically, he posited the following relationships: (1) Aggressive responses to the original stimulus generalize to similar stimuli, with the amount of generalization becoming smaller for increasingly dissimilar stimuli.

(2) Responses that conflict with aggressive reactions to the original stimulus also generalize to similar stimuli, becoming weaker for increasingly dissimilar stimuli.

(3) The gradient of generalization of inhibitory, conflicting responses falls off more steeply with stimulus dissimilarity than does that of aggressive responses.

(4) An increase in the drive associated with either gradient will raise the overall height of that gradient.

(5) Given alternative response avenues, the responses that will be executed are those associated with the greatest net strength – net strength being defined as the response strength of aggression minus the response strength of incompatible inhibitory reaction.

In its simplest form, Miller’s displacement model is expressed in intersecting linear gradients.

For linear generalization gradients, the strongest displacement is defined as the point at which the gradient associated with the strength of competing responses touches the abscissa. At this point, aggressive drive is unimpaired by inhibitory forces, and the response strength of aggression is above that for all other target points associated with negligible inhibitory forces.

Additionally, the net response strength of aggression – that is, the residual strength left after subtraction of the strength of competing responses – is above that of all target points associated with higher, yet impaired, aggressive drive.

The unique prediction, then, of Miller’s displacement model is that when aggression against the instigator is inhibited, aggression will be directed against the target (out of potentially many available targets) for which: (a) the strength of competing nonaggressive responses is negligible and (b) the stimulus similarity with the instigating target is the greatest. In this context, the inhibition of aggression is assumed to result from conflict, that is, from the dominant strength of aggression-opposing response tendencies. Miller distinguished this type of precondition for displacement from the prevention of aggression by the unavailability of the instigator as a target (Zillmann, 1979).

With regard to interpersonal human aggression, the following principal propositions inherent in Miller’s displacement model are significant: (1) The prevention or the inhibition of aggression against a particular instigator occasions aggression against other persons.

(2) The unavailability of the aggression instigator as a target for attack leads to aggression against highly similar persons. Relative to the strength of the attack that was prevented, the strength of the attack upon the highly similar substitute person is only trivially reduced.

(3) Given that aggression against the instigator is inhibited by conflicting response tendencies, aggression is displaced against persons of greater similarity to the instigator, the higher the aggressive drive and/or the lower the inhibitory forces. Conversely, aggression is displaced against persons of greater dissimilarity to the instigator, the lower the aggressive drive and/or the higher the inhibitory forces. Although the strength of the displaced attack is not a simple function of similarity between the instigator and the substitute, attacks upon persons who are highly dissimilar to the instigator tend to be of substantially lower intensity than the potential strength of the inhibited attack on the instigator.

The displacement model thus projects a rather gloomy view of efforts to control interpersonal aggression. It would appear that the control of aggression would best be served by letting the aggressively instigated person attack his or her tormentor. (Such an attack might at least be morally warranted.) Clearly, if this retaliatory attack does not occur, for whatever reason, aggression will victimize someone who is innocent. And since it seems inconceivable to control the instigated person’s environment (to prevent him or her from being provided with substitute targets), it can only be hoped that target similarity will be minimal, so that drive generalization will be weak (Zillmann, 1979).

All these implications of Miller’s model hinge on one crucial assumption: that inhibitory responses generalize less than aggressive ones. Miller derived this assumption from findings on conflict. Specifically, he drew upon the demonstration (Miller, 1944) that if a hungry animal is first trained to find food at the end of an alley and then given electric shock there, the tendency to avoid this location is weakened by distance more than the approach to food that it inhibits. The spatial gradient of avoidance is thus steeper than that of approach, and Miller felt that stimulus dissimilarity should function in a manner analogous to spatial separation. As far as interpersonal aggression is concerned, this suggestion is still entirely without empirical foundation, however. The lack of critical testing of this particular proposal is mainly due to what proved to be the major inadequacy of the displacement model: its vagueness with regard to the operationalization of stimulus similarity. In the transformation of the conflict model, the precision of the variable of spatial separation – that is, of physical distance – was lost, and the measurement of ‘stimulus similarity’ posed seemingly insurmountable problems.

The stimulus-response nomenclature leads one to expect a degree of precision in the similarity variation far beyond that which has been accomplished. Miller himself (e.g., 1948) invited unrestrained use of the similarity construct by espousing a form of semantic generalization. He proposed that “if the individual learns to respond to two quite different situations with the same verbal response, the stimuli produced by this response will be a common element mediating an increased amount of generalization from one situation to the other” (Miller, 1948). The process was referred to as the acquired equivalence of cues (Miller & Dollard, 1941), and it was considered to produce ‘secondary generalization’ (cf.

Hull, 1943). In other words, the generalization of aggressive responses, or of competing reactions for that matter, no longer was restricted to perceptual similarities based on the

physical manifestations of original and substitute targets, but it could bridge vast perceptual discrepancies as long as there were common verbal or nonverbal denoting or connoting qualifiers. To use one of Miller’s (1948) illustrations, the mechanism of secondary generalization permits Mr. Johnson, when forced to inhibit aggression against Mr. Bartlett, to displace his attack on Mr. Bartlett’s dog.

It should not be surprising, then, to find stimulus similarity experimentally manipulated along ‘semantic’ dimensions rather than perceptual ones (e.g., Murray, 1954; Murray & Berkun, 1955; Ferson, 1958).

Although perceptually obtrusive racial characteristics were considered important in the so-called scapegoat theory of prejudice (cf. Berkowitz, 1962), similarity has generally been derived from conceptual criteria. Assessments have been made on intuitive grounds, leaving room for considerable interpretative variation (Zillmann, 1979).

Revisions of Frustration-Aggression Theory There have been several attempts to modify frustration-aggression theory in order to accomodate some of the emerging findings. For example, Barker, Dembo & Lewin (1941) proposed a frustration-regression hypothesis. These investigators had observed a tendency for frustrated children to display ‘primitive’ behavior patterns. The children’s behavior gave the impression of a regression back to an earlier developmental stage. The reformulation simply emphasized this apparent regression without, however, changing the frustration-aggression relationship specified in the original hypothesis. Similarly, the basic relation between frustration and aggression was not changed by Maier (1949), who proposed a frustration-fixation hypothesis. This hypothesis was developed to deal with the finding that frustrated rats develop a tendency to perform a noninstrumental response very persistently – such as turning to a particular side. Clearly, these are minor, negligible modifications.

An attempt to revise frustration-aggression theory in a more sweeping manner was made by Berkowitz (1962, 1965a, 1969a). Berkowitz (1965a) proposed: (a) that frustration induces an emotional reaction – anger – that “creates only a readiness for aggressive acts”; and (b) that “aggressive responses will not occur, even given this readiness, unless there are suitable cues, stimuli associated with the present or previous anger instigators. Objects having some connection with aggression generally may also have this cue property”. Furthermore, Berkowitz sought to restrict the universal claim that all aggression presupposes frustration by suggesting that in the absence of frustration, exposure to suitable cues can lead to the formation and evocation of aggressive habits.

Except for the latter qualification, Berkowitz’s revision to a large extent rephrases frustration-aggression theory as amended by Miller (1941, 1948). The earlier concepts of aggressive drive and response strength for aggression have been replaced by those of emotional anger and aggression readiness, and the stimulus dependence, which was stressed

by Miller, is similarly emphasized. It is the involvement of the concept of ‘suitable cues’ then, that constitutes a departure from the earlier model (Zillmann, 1979).

In Berkowitz’s revision, this concept of ‘suitable cues’ is central, yet highly ambiguous. The definition of these cues (i.e., stimuli associated with present or previous anger instigators or with aggression generally) says nothing about the minimal strength of ‘association’ necessary to make a cue a suitable one. If matters are taken literally, all stimuli connected with the frustrating experience as such and with aggression in general become potent aggressive cues that are sufficient to evoke aggression in the frustrated individual. In every immediate social encounter that involves the frustration of one person by another, suitable cues are thus undoubtedly abundant. Such encounters, then, should always lead to aggressive exchanges –

an implication that runs counter to the objective of the revision. Furthermore, if any aggression-linked stimulus, as proposed, indeed suffices as a cue in evoking aggressive reactions, one would assume that the frustrated person will be induced to aggress against some available person or object unless he or she is placed in a highly artificial, restrictive environment. This is to say that the standard human environment is so laden with cues that somehow relate to hostility and aggression that it becomes difficult to see how any frustrated person can avoid cues suitable for the evocation of aggression. Incidentally, this line of reasoning shows that Berkowitz’s notion of suitable cues can readily accomodate the displacement of aggression. Where Miller’s model required target similarity or cue equivalence, Berkowitz’s notion calls for suitable cues – a seeming, rather than actual, difference from Miller’s model because similarity or equivalence assures a degree of association between targets and, consequently, the suitability of cues (Zillmann, 1979).

Berkowitz’s insistence on the involvement of suitable cues in the evocation of aggression is not to be confused with the notion of the availability of appropriate targets or the possible suggestive power of the availability of aggressive means. The revision does not so much suggest that a person who is motivated to aggress against his or her instigator will attack as soon as encountering both the target and the necessary means as it stresses that aggression can be ‘pulled out’ of the generally motivated and possibly even unmotivated aggressor. The individual is thus partly driven (by a state of anger) and partly pulled (by the environment) into violent action. Berkowitz (1965a) stipulated that the strength of an attack made in response to suitable cues is a joint function of: (a) “the aggressive cue value of this stimulus – the strength of the association between the eliciting stimulus and the past or present determinants of aggression”, and (b) “the degree of aggression readiness – anger intensity or strength of the aggressiveness habits”.

Although Berkowitz (1969b) has been a vehement critic of Lorenz’s instinct theory of aggression, the model just outlined obviously shares some principal features with it. Both approaches assign an elicitation power to certain stimuli. In Lorenz’s model, the connection between releaser and aggressive response is characteristically assumed to be innate. In Berkowitz’s model, this connection between suitable cues and aggressive response is generally assumed to be learned. The latter model thus seems to accomodate the human capacity for learning, which is largely ignored in the former. In spite of this difference, however, both models lead to the expectation that the organism is ‘forced into’ aggression when certain key stimuli are encountered. This evocation of aggression by key stimuli is seen as relatively independent of the level of aggression instigation; in fact, according to Berkowitz (e.g. 1970, 1973b, 1974), it can occur in the absence of annoyance. The person is, so to speak, surprised by a salient suitable cue and, as in Lorenz’s model, cannot help but launch an attack. In Berkowitz’s terms, the attack is ‘pulled out of’ the organism.

The pulling analogy would acceptably describe the process of the evocation of aggression if the aggressive reaction triggered were strictly interpreted as a conditioned response. Nearly any initially neutral stimulus can, no doubt, be made to trigger an aggressive response by the proper application of conditioning procedures. The analogy is not restricted to conditioning, however. Berkowitz suggested that the mere association of stimuli with aggression may provide suitable cues – that is, cues that pull aggression out of the organism. A knife or club, for example, can hardly be considered a stimulus to which specific aggressive reactions have been conditioned. Such potential weapons are nonetheless considered suitable cues for the evocation of aggression, especially in frustrated persons. The exact process assumed to mediate this proposed effect has not been made entirely clear (Zillmann, 1979).

According to Berkowitz’s reasoning, weapons generally are highly suitable cues in the elicitation of aggression. In somewhat dramatic language, he suggested: “Guns not only permit violence, they can stimulate it as well. The finger pulls the trigger, but the trigger may

also be pulling the finger” (Berkowitz, 1968). The frustrated person’s exposure to such cues as weapons should thus constitute conditions for a direct test of the effect of aggressive cues.

Berkowitz & LePage (1967) have, in fact, used this paradigm.

It was found that in nonangry subjects, the presence of weapons and their association with the aggression target had no discernible effect on aggressive behavior as measured in the number of electric shocks given to the confederate. In angry subjects, by contrast, the variation in cues resulted in differences in aggressiveness. Angry subjects behaved significantly more aggressively in the presence of weapons associated with the target than in the presence of the badminton rackets or in the absence of either set of gear.

In an attempt to extend the research concerning what has come to be known as the ‘weapons effect’, Ellis, Weinir & Miller (1971) failed to replicate the findings reported by Berkowitz & LePage.

Buss, Booker & Buss (1972) also reported unsuccessful attempts to replicate the weapons effect. The presence of weapons reduced rather than increased aggressive behavior, and it did so equally for instigated and noninstigated subjects. Page & Scheidt (1971) approached the elusive weapons effect in terms of experimental artifact. They proposed a “complex interplay between sophistication, demand awareness, and evaluation apprehension” in the experimental investigation of aggressive behavior.

In a recent field experiment on the effects of weapons on aggression, Turner, Layton & Simons (1975) bypassed the awareness problems of laboratory research. These investigators frustrated car drivers by having a confederate block traffic. The extent of horn honking of the driver in the car immediately behind the confederate was taken to reflect aggressive behavior.

Is horn honking an acceptable index of aggression ? Turner, Layton & Simons presented data that show that drivers consider themselves to be readily annoyed by the behavior of nonalert and discourteous fellow drivers. These drivers also reported little reluctance to express their annoyance. The question, however, is whether horn honking is more than such an expression of annoyance. There are no correlational data that show that a person who honks is out to inflict real harm or injury. The validity of the measure thus rests on a construct, and this construct appears questionable. Honking is commonly practiced by drivers, whether they are annoyed or not, when they realize that the person ahead has missed the signal change – presumably because of distraction rather than malice. Honking may well be a prosocial communicative act that merely tells the other driver that he or she can move on. At times, however, it may be employed to express anger, and we thus treat horn honking as a measure of the expression of annoyance. Since there is no apparent infliction of harm or injury involved, we cannot consider it a measure of hostile or aggressive behavior (Zillmann, 1979).

As it stands, the available research evidence concerning ‘aggressive cues’ such as weapons is largely negative, inconsistent, and contradictory (Zillmann, 1979). Regarding the degree to which cues are associated with the instigation of aggression and are correspondingly capable of evoking aggressive reactions, the evidence (Berkowitz, 1965b; Berkowitz & Geen, 1966; Geen & Berkowitz, 1966; Berkowitz & Geen, 1967; Geen & Berkowitz, 1967) is similarly equivocal.

As might be expected, the mediation paradigm has also been employed in the investigation of the displacement of hostility: e.g., Berkowitz & Knurek (1969).

The findings of Fenigstein & Buss (1974) show no discernible effect of a person’s association with an annoyer on the displacement of aggression. In their study, the friends of an annoyer simply did not constitute more suitable targets for aggression than other people. Needless to say, these findings also fail to support Miller’s displacement theory.

In summary, then, the overall evidence concerning Berkowitz’s mediation or association model is not unequivocal. The findings that are the least suspect of being confounded with demands and apprehensions are not supportive (Zillmann, 1979).

The validity of F-A Theory The fundamental assumption of F-A theory was stated categorically by Dollard et al. (1939): “Aggression is always a consequence of frustration”. From this statement it is not clear whether frustration is declared to be a necessary, or a sufficient, or both a necessary and sufficient condition of aggression. Clearly, the implications of the three conditions are different. If frustration is a sufficient condition but not a necessary one, we must expect that aggression always follows frustration but may also occur otherwise. If frustration is a necessary but not a sufficient condition, then aggression cannot occur unless first frustration occurs but need not occur even after frustration. If frustration is a necessary and sufficient condition, then aggression occurs if and only if frustration occurs. Further, “From the point of view of daily observation, it does not seem reasonable to assume that aggressive behavior of the usually recognized varieties is always traceable to and produced by some form of frustration”.

This seems to cast doubt on the assumption that frustration is a necessary condition for aggression. The argument that follows is not a defense of necessity but of sufficiency: “In many adults and even children, frustration may be followed so promptly by apparent acceptance of the situation and readjustment thereto that one looks in vain for the relatively gross criteria ordinarily thought of as characterizing aggressiveness... However the absence of overt aggressive reactions does not mean that such reaction tendencies are thereby annihilated”.

In other words, appearances are deceptive. Even though, following frustration, aggression is not always observed, it may nevertheless be there as a ‘tendency’. To put it another way, if one sticks with the assumption that “aggression is always a consequence of frustration”, one can always find evidence for it if one looks long enough. A sufficiently diligent search is bound to uncover aggression following frustration and (one presumes) a frustration to which aggression can be traced (Rapoport, 1974).

Commonly, young children are not able to gratify all their desires the moment they occur and are not able to derive satisfaction from anticipating deliberately postponed gratification.

Consequently, frustration is inevitable in every child’s life. It is inherent in the socialization process. The assumption that frustration is a necessary pre-condition of aggression cannot be falsified, simply because instances of frustration can always be found. Next, it is known that people learn to act in ways sanctioned by their society. Consequently, aggressive reactions must often be suppressed or directed into socially acceptable channels, perhaps in disguised form. If so then practically every action can be interpreted as a ‘tendency’ toward ‘aggression’, perhaps sublimated. Thereby the assumption that frustration is a sufficient condition for aggression cannot be falsified, simply because ‘tendencies’ toward aggression can always be found, or labeled as such. A theory that cannot be falsified is scientifically worthless since it does not distinguish between supporting and refuting evidence (Rapoport, 1974).

Anecdotal material supporting the frustration-aggression theory is, of course, abundant. A young man who is known to have hated his father and brother becomes an energetic political reformer, which provides opportunity to attack corrupt politicians, callous power figures, etc.

Thus, he demonstrates both a frustration-aggression pattern and the ‘displacement’ of aggression into channels acceptable, at least to his followers. Instances of this sort may well

corroborate the theory, but they are not evidence in the scientific sense of the word. To serve as evidence, such instances must be compared with instances where young men known to have ‘hated’ their parents did not become social reformers, and also with instances where young people who did not ‘hate’ their parents became social reformers. Such standards of evidence are often declared to be impossible to satisfy because it is part of the theory that aggression takes on many different forms and stems from many different sources. If so, it becomes all the more facile to relate ‘aggression’ to ‘frustration’ in practically any situation, and we are again left with an unfalsifiable hypothesis (Rapoport, 1974).

Several anthropological and sociological observations have been interpreted in accordance with the frustration-aggression theory. People in different cultural environments behave toward each other in different ways. In some societies there is a great deal of interpersonal, overt violence; in others, overt violence is practically unknown. A common-sense view would ascribe more aggressiveness to the people in the former society and less to those in the latter.

If, however, one insists that frustration is the common lot of all children in the process of socialization, one is tempted to look for other manifestations of hostility if fighting is not observed. One might notice that in the non-fighting society there are more verbal assaults, quarreling, ‘malicious gossip’, etc. Similarly, among tribal societies, some tribes are warlike, others are peaceful. Ordinarily one might suppose that the former are more aggressive than the latter. If, however, one allows that there is in all people a reservoir of hostility accumulated in the process of socialization, one will interpret war-making as hostility directed outward and will conclude that the non-warlike people direct hostility inward towards members of their own tribes or against themselves. Such interpretations are not necessarily false, but they need to be supported by ‘harder’ data if they are to be taken seriously. For example, in finding that high murder rates are associated with low suicide rates, and vice versa, if it can be shown that the correlations are unlikely to be coincidental, then a case can be made for the notion that aggression will be manifested in one way or another, and that if one way is blocked, the other will be taken. Still, the relation between frustration and aggression remains an elusive one.

In the sociological sphere, the frustration-aggression theory has been used to explain criminality. Innumerable studies have revealed correlations between the incidence of criminality and low economic status. Bonger (1916) cites Hausner’s finding that during peace time the criminality of German soldiers is 25 times that of civilians. It is noted that during the first two years of the Nazi regime, the incidence of murders in Germany increased while thefts and burglaries decreased. It is impossible to decide whether these findings are in any meaningful way related to the frustration-aggression theory. It can, and has been, argued that poverty is a source of frustration; so is the strict discipline of the barracks, with which the Nazi regime is also associated. Nevertheless, the same data may be used against the theory with equal effectiveness, Let us say, as is commonly maintained, that children of the lower classes are subjected to fewer frustrations than those of the middle and upper classes; that soldiers, being trained to focus their hostility on the ‘enemy’ (real or hypothetical), might be expected on that basis to manifest less (not more) aggression (criminality) in their own society. Then the same can be said of the Germans during the Nazi regime.

The most serious flaw in relating these observations to the frustration-aggression theory appears to be, as before, the vagueness of the terms. What is criminality? And is it necessarily related to aggression? A poor man is more likely to steal (in the accepted sense of the word) than a rich man. Does this mean that the poor man is more aggressive? On the other hand, the rich man is more likely to engage in socially approved aggressive acts (say, large-scale financial operations) than the poor man. Is this an indication that the rich man is more frustrated than the poor man? Or can it be that aggression is as likely, or more likely, to be

instigated by success than by frustration ? Recall the mouse made pugnacious by the simple expedient of ‘fixing’ all his fights.

The most pronounced impact of the frustration-aggression theory has been on child rearing.

Attention was called to the vicious cycle: initial frustration (unavoidable) – direct aggression – inhibition of aggression, resulting in more frustration – displaced aggression, often inducing neurosis. It was recommended that the cycle be broken at the inhibition stage, and permissiveness became a key word in the lexicon of progressive parenthood. What difference the change in child-rearing practices has made in the generations since grown up is hard to say. We are told that the present unrest among the young (frequently equated with ‘aggression’) is a direct consequence of exaggerated permissiveness. It is equally plausible that the aggressiveness of the discontented young (if it is aggressiveness) is a result of intense frustrations related not to child-rearing practices but to a feeling of helplessness in the face of maniacal government policies and festering social evils.

In summary, although selected instances showing linkages between clearly defined frustration and clearly defined aggressive reactions (direct or displaced) are impressive, the lack of explicit criteria for recognizing more general manifestations of either frustration or aggression makes it impossible to evaluate the extent to which the frustration-aggression theory is valid (Rapoport, 1974).

Criticism of the Concept of Catharsis (Kaufmann, 1965) The term ‘catharsis’ in relation to the F-A hypothesis, was originally used to denote an aggressive act followed by lowered subsequent aggression. This lowering of aggressive drive or, in terms of Miller’s (1941) emendation, of instigation to aggression was acribed to a draining away of drive leading to aggression. Rosenbaum & DeCharms (1960) while studying the cathartic effects of vicarious aggression, found that the mere opportunity to respond to a frustrater lowered subsequent derogation of him, even where the original response was not an aggressive one (though this effect was present for low self-esteem subjects only). The issue of catharsis and the cathartic effect is further confused by conflicting experimental findings.

Aggressive behavior has been found to lower subsequent aggression in some instances (Feshbach, 1955; Thibaut, 1950; Thibaut & Coules, I952) but in others has produced an increase in aggression (DeCharms & Wilkins, 1963; Kenny, 1953).

Feshbach himself (1956) obtained such findings in a later experiment and presented an interesting analysis of the aggression-enhancing or reducing effects of (aggressive) fantasy behavior. He hypothesized that aggression is cathartic where the subsequent situation is related to the situation involving the aggressive act, but that a second opportunity to aggress should lead to enhanced aggression where the two situations are unrelated.

Stagner (1944) argued that where aggression is an outcome of habitual hostility no cathartic effect is to be expected; presumably, since greater habit strength develops with repeated aggression, the strength of the latter should increase on those grounds. According to Buss (1961) catharsis should decrease subsequent aggression where anger is present but increase it in the absence of anger because in the latter case inhibition has been lowered.

DeCharms & Wilkins (1963) presented strong evidence that hostile statements engender subsequent heightened aggression and at the same time clarified an issue raised by Worchel (1960) dealing with the instigation-reducing effects of status restoration. In Worchel’s experiment, the attacker and the status restorer were one and the same person, which made it impossible to separate the effects of status restoration from those of attacker’s changed image as a supportive status restorer. From the Rosenbaum & DeCharms (1960) results it might be inferred that status equalization, that is, reduction of the attacker’s status, may serve as a

substitute for victim’s status restoration (see also Feshbach, 1964). DeCharms& Wilkins (1963) found that status restoration by the subject himself or by a vicar was positively related to subsequent aggression, rather than negatively, as had been found by Worchel (1960).

Again, one is tempted to infer that the social connotations of retaliation against a (deviant) attacker outweighed emotional aspects of accumulated anger or anxiety.

The evidence is thus truly conflicting. Aggression enhances or reduces subsequent aggression; a vicar’s aggression may serve either as a substitute (Rosenbaum & DeCharms, 1960), or as a model (DeCharms & Wilkins, 1963) for the subject’s subsequent aggression.

However, the difficulties inherent in the interpretation of catharsis phenomena do not end here. After all, the cathartic effect may be due to any of the following reasons: (a) the subject has ‘let off steam’ (an undifferentiated autonomic arousal model), (b) the subject has aggressed against someone or other (a hydraulic aggression model), (c) the subject has successfully aggressed against an indeterminate target (an expressive model), (d) the subject has successfully aggressed against the instigator (a self-esteem restoration or threat removal model), (e) the subject has seen the frustrater punished, even by someone else, and cognitive expectations regarding justice have been confirmed (a dissonance reduction model), (f) the subject has accumulated reactive inhibition or fatigue (a behavioral oscillation model), (g) the subject has developed guilt or aggression anxiety subsequent to aggression (an aggression-anxiety-with-delay model), (h) the subject perceives that a given amount of aggression is appropriate under the circumstances (a social reality model).

Where a cathartic effect is ostensibly absent, it may be the case that his aggressive habits have become stronger, that he perceives greater social desirability or tolerance for aggression, or that he perceives aggression to be the most effective way to deal with the situation, etc.

A series of experiments by Hokanson (1961) and his collaborators allows for interesting inferences regarding the role of expectations in determining the need for catharsis in order to produce the cathartic effect. Hokanson & Burgess (1962), for example, found that aggression against a high-status attacker did not reduce sympathetic arousal, while aggression against a low-status attacker did. These results appear to be directly relevant to the above mentioned variable of perceived or expected effectiveness of the attack. Hokanson & Shetler (1963) found that when the subject is attacked by a high-status individual and does not expect an opportunity for retaliation his systolic blood pressure subsides at the same rate as when he was able to aggress. Though the authors discussed these results mainly in terms of substitute responses to frustration, such as withdrawal, an interpretation along the lines of perceived and – in the case of the high-status frustrater – accepted social expectation appears quite plausible.

In any event, Berkowitz’ (1962) exception to the absence of clear-cut interpretations to current experiments was well taken; too many variables have remained uncontrolled for any verdict regarding the notion of catharsis to remain undisputed or even to be meaningful.

Criticisms of F-A theory Frustration-aggression theory soon became accepted by a majority of social psychologists and was later taken up by other social scientists who used its basic assumptions in the development of the influential deprivation theory of violence and aggression.

Dollard et al.’s frustration-aggression hypotheses were devastatingly simple: (1) “aggressive behavior always presupposes the existence of frustration”, and (2) “the existence of frustration always leads to some form of aggression”. Miller (1941), one of Dollard’s co-authors, later retracted the second part of the theory on the grounds that aggression was only one of a whole range of alternative responses that could be triggered by

frustration. However, there are other possible objections to the initial hypothesis. Acts of promiscuous cruelty and sadism, or what Fromm (1974) terms ‘malignant aggression’, are by no means always instigated by frustration. And many of the more spontaneous aggressive acts may be responses to noxious stimuli rather than to frustration. Frustration is a confusing and ambiguous term. It can mean simply denial or deprivation of some ‘good’ desired by an individual or group; but this usage is misleading because it implies that there is some psychological state in common between those who are deprived of or denied different kinds of desired ‘good’. But there is clearly a world of difference between the feelings of, say, a nationalist who finds the demand for national independence rejected and those of a motorist who is deprived of gasoline by a tanker-drivers’ strike. It is not clear what there is to be gained by the use of such undiscriminating terms in theory building. Dollard’s frustration-aggression syndrome may be more relevant to the somewhat rarer type of frustration experienced when an on-going goal-directed activity is interrupted. But, even among members of the same family or community, responses to denials or deprivations of the same desired good vary enormously in nature and intensity. As Fromm (1974) argues, the kind of frustration an individual experiences, and the likelihood and nature of any aggressive behavior induced thereby, will depend considerably on individual personality differences. In a crude way, when we recognize wide differences in individual capacities for patience and tolerance, we are admitting the empirical variability of individual thresholds for aggressive responses (Wilkinson, 1979).

A further basic objection against frustration-aggression theory is its imprecision in the use of the concept of aggression.

Even if we accept that aggressive behavior is frequently preceded by frustration (without for a minute conceding that frustration always leads to aggression), surely it is important to know what forms and intensities of frustration are linked to what forms and intensities of aggression. In the context of political violence, we need to know what particular forms and intensities of collectively perceived deprivation lead to riot, rebellion, terrorism, revolution, and civil war. And why is it that large-scale collective deprivations (e.g., those of famine, economic collapse, or defeat in war) sometimes lead to massive acts of political violence, yet sometimes fail to do so? Frustration-aggression theory as such has not seriously grappled with these problems. However, relative deprivation theory, which derives some of its basic assumptions from frustration-aggression theorists such as Dollard and Berkowitz, does appear to offer concepts and insights of more direct applicability to political violence.

The insights that popular satisfactions and discontents are of a relative nature, and that feelings of injustice can be intensified by a revolution of rising expectations or exacerbated by a sudden crisis, were familiar to earlier generations of social analysts. Marx (“Wage Labour and Capital”, 1847) notes: “Our desires and pleasures spring from society; we measure them, therefore, by society and not by the objects which serve for their satisfaction. Because they are of a social nature, they are of a relative nature”. And there is a vivid perception in Durkheim (1893) of the extent to which the hierarchic distribution of rewards of society is sanctioned and legitimated by moral beliefs.

Clearly, it is an important truth that low levels of poverty or deprivation do not automatically breed responses of aggression and violence. Hence, the poorest masses of the Third World are unavailable and unprepared for revolutionary mobilization or indeed any political participation: they are simply engaged in a daily struggle to keep alive. They lack both energy and resources to organize political parties or movements of their own. Envy, on the other hand, is universal, though its sociology has been almost totally neglected. Hence, even in an affluent community in which all the population share to some degree in the prosperity, there

will be those who covet their neighbors’ goods or who envy others for their knowledge, skills, status, or political power.

Modern theorists of the relative deprivation school of aggression have been preoccupied with exploring the implications of frustration-aggression theory for the analysis of civil violence.

According to this theory, at the individual level, perception of frustration arouses anger, which then functions as a drive (Berkowitz, 1965). Gurr (1968) argues that the implication of this theory is “that civil violence almost always has a strong appetitive sensational base and that the magnitude of its effects on the social system is substantially dependent on how widespread and intense anger is among those it mobilizes... If anger implies the presence of frustration, there is compelling evidence that frustration is all but universally characteristic of participants in civil strife”.

Gurr’s basic premise is that the necessary precondition for violent civil conflict is relative deprivation (RD), defined as the actors’ perception of discrepancy between their value expectations and their environment’s value capabilities. In Why Men Rebel (1970), Gurr defines RD as the perceived discrepancy between men’s value expectations and their value capabilities. “Value expectations are the goods and conditions of life to which people believe they are rightfully entitled. Value capabilities are the goods and conditions they think they are capable of getting and keeping”. He uses a threefold classification of types of values: welfare, power, and interpersonal. Interpersonal values are further subdivided into status, communality, and ideational coherence, but this does not help us very much because Gurr does not adequately explain these terms. Gurr’s axioms are interesting but rather too loosely drawn to offer any theoretical purchase on the problem of political violence. For example: “the more intensely people are motivated toward a goal or committed to an attained level of values, the more sharply is interference resented and the greater the consequent instigation to aggression”. The term ‘instigation’ is itself unsatisfactory because it is not clear whether Gurr means the goal-seeker will simply be further incited to civil violence or whether violence will actually be brought about. Moreover, surely the nature of the goals and values involved is crucially important. What if the goal-seekers are religious or pacifistic? May they not interpret the interference as an incitement to turn the other cheek? Moreover, this ‘axiom’ illustrates another grave flaw in Gurr’s relative deprivation thoery of ‘Why Men Rebel’. Gurr constantly generalizes about intensities of motivation, anger, and resentment and about likelihoods and magnitudes of civil violence; yet nowhere does he explain how these things can be scientifically measured (Wilkinson, 1979). Some of Gurr’s generalizations have a surface appearance of common sense until one begins to test them against historical realities.

For example: “The likelihood and magnitude of civil violence tend to vary inversely with the availability of institutional mechanisms that permit the expression of nonviolent hostility”.

How can that assertion stand up in the light of the American Civil War, or the endemic civil violence in postcolonial India, or the intensity of violence in Northern Ireland? We should note, also, that Gurr does not explain how or why some aggression is mediated into actual violence. For this reason, Why Men Rebel can be more accurately described as the presentation of a model rather than a theory.

Finally, and perhaps most damaging of all, Gurr’s theory does not allow for the phenomena of elite-initiated and elite-directed violence in which mass participants are under authoritarian, military or paramilitary control. In such circumstances, by no means unfamiliar in internal wars, mass participants’ perceptions of relative deprivation are utterly irrelevant to explaining the causes of the violence. Gurr’s point that elite groups themselves sometimes resort to aggression as a consequence of the frustration-aggression syndrome is a fair one. However, it would be a distorted view of history that saw all aggressive wars as being induced by elite frustration and omitted all consideration of wars as acts of policy or as expressions of dynastic or national self-assertion (Wilkinson, 1979). In this analysis, criticisms of relative deprivation

theory of violence have thus far been directed at the work of Gurr. This is to some extent justified by the fact that he is the most prolific and widely known exponent of this theoretical approach. His trenchant remarks on everything from revolution and riot to the crime rate are avidly noted by commissions and experts. Nevertheless, there are other influential contributors to this school who have explored more fully than Gurr the political implications of four basic models of relative deprivation of social groups, classes, or societies: (1) the revolution of rising expectations in which social expectations and aspirations rise at a much faster rate than capabilities; (2) the so-called J-curve situation when capabilities increase and keep pace with rising expectations for a short period and then suddenly suffer a sharp decline; (3) the serious crisis or malaise that causes a reduction in capabilities while expectations remain constant; and (4) situations in which expectations and aspirations increase while value capabilities remain constant.

It is, of course, true that none of these models was invented by contemporary relative deprivations theorists. All are deployed and discussed in the somewhat more elegant literature of nineteenth century social theory. For example, Alexis de Tocqueville (1856) delineated situations (1) and (2) as contributory causes of the French Revolution. He brilliantly discerned the crucial role of changing expectations in the prerevolutionary period.

“The evils which were endured with patience so long as they were inevitable seem intolerable as soon as a hope can be entertained of escaping from them. The abuses which are removed seem to lay bare those which remain, and to render the sense of them more acute; the evil has decreased, it is true, but the perception of the evil is more keen. Feudalism in all its strength had not inspired as much aversion in the French as it did on the eve of its disappearance”.

De Tocqueville, however, was careful not to elevate his insights into a monocausal theory of revolution. In comparison, the confident generalizations of Davies (1969), the contemporary exponent of J-curve theory of relative deprivation, are based on a narrow and dogmatic reductionism.

“Revolution is most likely to take place when a prolonged period of rising expectations and rising gratifications is followed by a short period of sharp reversal, during which the gap between expectations and gratifications quickly widens and becomes intolerable. The frustration becomes focussed on the government, the violence becomes coherent and directional. If the frustration is sufficiently widespread, intense, and focussed on the government, the violence will become a revolution...”.

Marx and Engels made the most influential classic contribution to exploring the implications of model (3) in their theory of the roles of major economic crises and the gradual immiserization of the proletariat as contributory causes of revolution and civil conflict. But again their theories of revolution are infinitely more subtle and complex than the so-called empirical theories of the modern relative deprivation school. Marx and Engels went beyond the evidence when they claimed that differences in their ideologies, beliefs, and expectations of the various social classes are invariably a reflection of their relationship to the means of production. Nevertheless, Marxist theory of revolution does possess a historical sweep and a wealth of insights that modern relative deprivation theory cannot begin to match. Marxist theory does take into account the phenomenon of the unawakened exploited class that has not yet recognized either the injustice of its situation or its revolutionary potential. Marxist theory therefore lays great emphasis on the revolutionary movement’s responsibility for equipping

the proletariat with revolutionary consciousness, leadership, and the correct revolutionary theory. It can be argued that although Marx made somewhat crude and unsupported assumptions about the nature of bourgeois rationality, he did not make the mistake of assuming that the working-class masses were all efficient maximizers who would be able, unaided as it were, to arrive at an adequate understanding of the nature and causes of their oppression and the means of their emancipation. In the last resort, Marx’s revolutionary theory does not rely on spontaneity or the natural evolution of revolutionary consciousness.

Marx and Engels were among the first social theorists to recognize the importance of popular ideologies and beliefs in legitimating, explaining, supporting, criticizing, or denouncing specific social orders and their various distributions of wealth, status, and power: they were thus saved, almost in spite of themselves, from falling into the trap of their own materialism (Wilkinson, 1979).

In Why Men Rebel, Gurr disparages the notion that ideologies ‘cause’ violence. Yet although ideologies qua abstract doctrine do not in themselves directly cause violence, ideological movements, which define enemies and incite to combat, do frequently instigate political violence, wars, and ‘crusades’, with or without benefit of widely perceived relative deprivation. The work of Gurr and his colleagues reveals a failure to appreciate the influence of certain ideologies in creating and legitimating aspirations and expectations and equipping militant minorities to act in the name of the masses. Ideologies both define and articulate the sense of deprivation and claims of injustice and oppression. Most important of all, for our purposes, they bring powerful influence to bear in determining the kind of response made to frustration and anger by key groups or social strata. If an ideology tends to sanction violence or even to encourage and glory in it, then this inevitably strongly influences the nature of its adherents’ collective response to anger, rejection, or opposition.

Relative deprivation theory, then, seriously underrates and misunderstands the role of ideology in inducing discontent, aggression, and conflict (Wilkinson, 1979). In searching for alternative general causes of civil violence, certain relative deprivation theorists have plumped heavily for socioeconomic modernization. Feierabend, Feierabend & Nesvold (1969, 1973) argue that the modernization process tends to simultaneously intensify modernizing groups’ aspirations while challenging the entrenched positions of traditional groups. Hence, the probability of conflicts between ‘modernizers’ and ‘conservatives’ is enhanced as socioeconomic development accelerates. The faster the rate of modernization, they conclude, the greater the degree of disruption and the conduciveness to civil violence. The Feierabend and Nesvold hypothesis is strongly reinforced by the work of Huntington (1968, 1971) on political violence in what he terms ‘transitional’ societies, societies that are in the midprocess of modernization. Huntington does not question the assumption that increased relative deprivation renders civil violence more probable. His analysis.suggests that rapid modernization always involves intensified relative deprivation because it widens the gap between the changing aspirations and capabilities of the groups involved: social mobilization, education, and increased opportunities of political participation enhance aspirations while the already inadequate levels of production, employment opportunities, and governmental and administrative resources cannot keep pace with fresh expectations and needs. For some groups, capabilities actually decline (for example, peasants who are made landless by agricultural modernization programs and who cannot find work in the cities). Huntington observes that deprivation of political capabilities (such as the basic civil liberties and the right to vote) is more likely to lead to civil violence than purely economic deprivation.

“By broadening the range of political participation, political modernization increases the scope of civil violence, unless broader institutional channels for peaceful participation also come into being. The combined effect of both social-economic and political modernization,

however, is to undermine traditional political institutions and to make it most difficult to create broad-based political party systems and other forms of political institutions. The idea of ‘peaceful change’ or ‘development without violence’ thus becomes almost wholly unreal” (Huntington, 1971).

The whole tone of Huntington’s discussion is heavy with gloomy realism. He argues that the modernization process even stimulates and intensifies traditionally-rooted communal conflicts and that most of the forms of political violence likely to ensue from development – praetorian violence, political repression, and communal conflict – are of a destructive and debilitating character. None of the four general models of relative deprivation discussed by Huntington allows for the possibility of declining expectations and aspirations. Why should this be considered empirically impossible? Is it so exceptional for individuals and groups to actually lower their expectations or throw aside long-term aspirations – especially in conditions of acute crisis such as those caused by natural disaster or war? Humanity in extremis tends to adapt by concentrating on the immediate problems of survival, clutching hold of families and possessions, and riding out the storm. The mass of the world’s poor suffer in passive resignation. Most are resigned to the fact that their conditions are more likely to worsen than to improve. When mass despair and fatalism are omitted from the academic theorist’s model of the world, one suspects that the theorist has himself fallen prey to one or other of the powerful ideologies of modernization championed by Western aid experts, nationalist elites, and communist revolutionaries, all of whom have their own versions of the gospel. It is a cardinal error to confuse elite ideologies of modernization with the actual ideologies, beliefs, hopes, and fears of the masses (Wilkinson, 1979).

Whiting (1944) suggested that the culturally relevant overt reactions to frustration may be divided into four classes of social behavior: aggression, submission, dependence, and avoidance. These four classes were found to be adequate to describe overt frustration-reactions in Kwoma culture, and it is proposed that they may be useful in the analysis of other cultures.

Owing to certain universal conditions of social life (it is suggested), aggression, submission, dependence, and avoidance will be likely to be specified as proper responses to frustration in given contexts in all societies. The analysis of Kwoma culture shows that the frustrationdependence sequence occurs both during infancy, when the infant is too immature to care for himself, and later in life as a part of cooperative reactions to common frustrations; that the frustration-submission reaction occurs in childhood during the process of socialization; that the frustration-avoidance reactions occur in children when they are frustrated by persons bigger and stronger than they, and in adults when they are frustrated by persons whose supporting group is more powerful than their own; and, finally, that the frustration-aggression reaction occurs in education and social control. Since education, social control, co-operation, and variation in the power of the supporting group may be assumed to be universal cultural aspects, and helplessness of infants and disparities in the size and strength of individuals are universal biological facts, it may be assumed that these four reactions will be likely to occur in all cultures (Whiting, 1944).

Various authors have used other systems of classifying reactions to frustration. Dollard et al.

(1939) and Sears (1941) have distinguished between instrumental and goal behavior. This distinction was made primarily on the psychological level whereas the present analysis is made on a cultural level, and does not contradict the above, but was found to be more useful for a cultural analysis. The nine-fold classification suggested by Allport, Bruner & Jandorf (1941) was developed to describe frustration under conditions of social change, while the present classification is intended to describe the types of reactions to daily life frustrations.

Rosenzweig (1938-39) classified reactions to frustration on the personality level as: adequate – inadequate, direct – indirect, defensive – perseverative, and specific – nonspecific.

Sabini & Silver (1982) have argued that the Yale-school (Dollard et al., 1939) dictum ‘frustration produces aggression’ has been confused with ‘transgression produces anger’. In their social and moral account of anger, Sabini & Silver state: “Frustration-aggression as an account of anger appears wanting; it doesn’t even mention the word ‘anger’. But this omission should not be taken too seriously. After all, the authors of this theory were old-fashioned behaviorists, too circumspect to be seen in the company of ‘anger’, a word tainted with mentalism. But in their behavior, if not in their language, it was anger, or angry reactions, that concerned them. We can tell it was anger they wanted to talk about by looking at what their account professed to explain, i.e., what kinds of aggression were left after they chipped away those kinds their account was not meant to explain. Strangely, they denied interest in the causes of: armed robbery, invasion of another country, killing your aunt to get her millions, or calmly but randomly murdering pedestrians because of a brain lesion – all things we might call aggression. But they did include in their story of ‘aggression’: screaming at someone who deliberately steps on your toe, punching a rapist harder than is necessary to ensure separation, or the desserts of my bullying. The kinds of acts their account excludes are all instances of aggression without anger; yet they include cases of anger without aggression (for example, those we’ve just mentioned). We conclude from this that it was really anger they talked about, or at least angry reactions” (Sabini & Silver, 1982).

For the behaviorist all moral notions were subjective – personal reactions with emotional connotations – not objective, verifiable claims that could be included in a proper scientific theory. This ‘fact’ about morality was not a discovery of empirical research, but a presupposition, a gift from the philosophy of the time – logical positivism.

If one takes the view that morality is subjective, then ‘aggression’, ‘violence’, and ‘hostility’ are synonyms; they differ only in connotation or the subjective emotional reaction they provoke, rather than in denotation, and the theorist is free to use whichever one pleases him to describe behavior. The same view of morality that gives the theorist value-free aggression settles the status of the moral story of anger.

The hydraulic theory, then, did not seriously deal with the moral side of anger because it found moral talk too disreputable to include in its theorizing. Berkowitz (1952) attempted to incorporate the moral story within the frustration-aggression frame. He claimed that frustration always leads to anger. Pastore (1952) pointed out that whether you became angry at a bus-driver who passed you by depended on whether the bus had an ‘Out of Service’ sign on it or not. Whether it did or not, it was equally frustrating, but if it did, then it was a justified frustration; if it didn’t, it was unjustified (he called it arbitrary). Pastore claimed that we became angry at ‘arbitrary’, but not justified frustrations. Berkowitz answered Pastore in the following way. First, following Pastore, he calls the distinction between unjustified and justified a distinction between the arbitrary and the justified. Then he takes arbitrary to mean unexpected. And, of course, unexpected frustrations are more frustrating. But how was the bus that was out of service more expected than the one that wasn’t? Next, Berkowitz decides that you really would be angry but not show it. This sometimes happens. But suppose you’re frustrated by someone who misses an appointment because his mother died. Couldn’t you just feel pity? Does he deny this possibility?

He also claims it is your interpretation of the frustration that counts. But this is rampantly circular. Then he claims that among the ongoing, goal-directed activities we are always engaged in are, ‘the internal responses oriented toward the preservation of security and comfort’ or ‘impulses of self assertion’ (1962, p. 30). For a behaviorist, he certainly does

multiply internal, mental entities. We find these ad hoc accounts shaggy. We recommend Ockam’s razor’ (Sabini & Silver, 1982).

Sabini & Silver (1982) argued that “The perception of wants frustrated by a transgression is what leads to anger”, and that the concept of transgression implies notions such as culture, shared norms, common understanding.